R/stopFunctions.R

In FMM: Rhythmic Patterns Modeling by FMM Models

################################################################################
# Functions to check the criterion convergence for the backfitting
#   algorithm.
# Functions:
#   alwaysFalse:  returns FALSE to force maxiter iterations.
#   R2: check the converge criterion based on explained variability.
################################################################################

################################################################################
# Internal function: FALSE
# Arguments:
#   vData: data to be fitted an FMM model.
#   pred: fitted values from current iteration.
#   prevPred: fitted values from previous iteration
# Returns FALSE to force maxiter iterations.
################################################################################
alwaysFalse <- function(vData, pred, prevPred){
  return(FALSE)
}

################################################################################
# Convergence function: to check if the convergence criterion based on the
#                    difference between the explained variability in two
#                    consecutive iterations, is reached.
# Arguments:
#   vData: data to be fitted an FMM model.
#   pred: fitted values from current iteration.
#   prevPred: fitted values from previous iteration.
#   difMax: smallest observable difference between iterations
#           to stop backfitting algorithm.
# Returns TRUE when the converge criterion is reached.
################################################################################
R2 <- function(vData,pred,prevPred,difMax = 0.001){
  usedStopFunction <- function(vData, pred, prevPred){
    prevR2 <- PV(vData, prevPred)
    R2 <- PV(vData, pred)
    R2dif <- R2 - prevR2
    return(R2dif < difMax)
  }
  return(usedStopFunction)
}